[CRIU] Re: [PATCH 5/6] parasite: Don't assign error number on logfd
Pavel Emelyanov
xemul at parallels.com
Tue Feb 14 10:15:58 EST 2012
On 02/14/2012 07:09 PM, Cyrill Gorcunov wrote:
> On Tue, Feb 14, 2012 at 06:48:42PM +0400, Pavel Emelyanov wrote:
>>>
>>> Not at all. How pray tell you will be debuggin stuff when
>>> someone send you bug-report with no output from parasite?
>>> Having NO output is a way worse than having SOME output.
>>
>> There will be NO some output with logfd == 2. If you fail to
>> send logfd to parasite, just abort the dump.
>>
>
> Before we send logfd to parasite the parasite is already
> injected and ready to work and when logfd flights in we can
> produce output inside parasite and I'm voting to add
> debug code here and there (either via refined pr_debug
> or anything else). So loosing ability to early-print
> is inacceptable for me.
That's fine, but try not to write these logs into the target task's
already-open descriptors :)
> Cyrill
>
More information about the CRIU
mailing list